Here are implementations of different types of filter banks:
MDCT (Modified Discrete Cosine Transform)
LDFB (Low Delay Filter Bank)
QMF (Pseudo Quadrature Mirror Filter bank)
The first link leads to fast implementations using Python: Fast Python Implementations
The next link leads to an implementaion, using Python Keras, which a Neural Network library: Python Keras Implementations
It is advantageous when implenting filter banks together with neural networks for deep learning. It also can take advantage of a computers GPU and speed up the computation by parallelization, whithout a dedicated fast implementation.